arXiv:1207.2084v1 [astro-ph.IM] 9 Jul 2012 GREGOR Fabry-P´ erot Interferometer – status report and prospects Klaus G. Puschmann a , Horst Balthasar a , Christian Beck b , Rohan E. Louis a , Emil Popow a , Thomas Seelemann c , Reiner Volkmer d , Manfred Woche a , and Carsten Denker a a Leibniz-Institut f¨ ur Astrophysik Potsdam, An der Sternwarte 16, 14882 Potsdam, Germany b Instituto de Astrof´ ısica de Canarias, c/ V´ ıa L´ actea s/n, 38205 La Laguna, Spain c LaVision, Anna-Vandenhoeck-Ring 19, 37081 G¨ ottingen, Germany d Kiepenheuer-Institut f¨ ur Sonnenphysik, Sch¨ oneckstraße 6, 79104 Freiburg, Germany ABSTRACT The GREGOR Fabry-P´ erot Interferometer (GFPI) is one of three first-light instruments of the German 1.5-meter GREGOR solar telescope at the Observatorio del Teide, Tenerife, Spain. The GFPI allows fast narrow-band imaging and post-factum image restoration. The retrieved physical parameters will be a fundamental building block for understanding the dynamic Sun and its magnetic field at spatial scales down to 50 km on the solar surface. The GFPI is a tunable dual-etalon system in a collimated mounting. It is designed for spectropolarimetric observations over the wavelength range from 530–860 nm with a theoretical spectral resolution of R ≈ 250, 000. The GFPI is equipped with a full-Stokes polarimeter. Large-format, high-cadence CCD detectors with powerful computer hard- and software enable the scanning of spectral lines in time spans equivalent to the evolution time of solar features. The field-of-view of 50 ′′ × 38 ′′ covers a significant fraction of the typical area of active regions. We present the main characteristics of the GFPI including advanced and automated calibration and observing procedures. We discuss improvements in the optical design of the instrument and show first observational results. Finally, we lay out first concrete ideas for the integration of a second FPI, the Blue Imaging Solar Spectrometer, which will explore the blue spectral region below 530 nm. Keywords: Sun — spectroscopy — polarimetry — high angular resolution — instrumentation — image restoration 1. INTRODUCTION Solar physics has made tremendous progress during recent years thanks to numerical simulations and high-resolution, spectropolarimetric observations with modern solar telescopes such as the Swedish Solar Telescope, 1 the Solar Optical Telescope on board the Japanese HINODE satellite, 2 and the stratospheric Sunrise telescope. 3 Taking the nature of sunspots as an example, many important new observational results have been found, e.g., details about the brightness of penumbral filaments, the Evershed flow, the dark-cored penumbral filaments, the net circular polarization, and the moving magnetic features in the sunspot moat. Telescopes with apertures of about 1.5 m such as the GREGOR solar telescope 4–7 or the New Solar Telescope 8, 9 will help to discriminate among competing sunspot models and to explain the energy balance of sunspots. New results on the emergence, evolution, and disappearance of magnetic flux at smallest scales can also be expected. However, these 1.5-meter telescopes are just the precursors of the next-generation solar telescopes, i.e., the Advanced Technology Solar Telescope 10 and the European Solar Telescope, 11 which will finally be able to resolve the fundamental scales of the solar photosphere, namely, the photon mean free path and the pressure scale height. Fabry-P´ erot interferometers (FPIs) have certainly gained importance in solar physics during the last decades, because they deliver high spatial and spectral resolution, and a growing number of such instruments is in operation at various tele- scopes. Although most of the instruments have been initially designed only for spectroscopy, most of them have now been upgraded to provide full-Stokes polarimetry. 12– 14 The Universit¨ ats-Sternwarte G¨ ottingen developed an imaging spectrom- eter for the German Vacuum Tower Telescope (VTT) in the early 1990s. This instrument used a universal birefringent filter (UBF) as an order-sorting filter for a narrow-band FPI mounted in the collimated light beam. 15 The spectrometer was later equipped with a Stokes- V polarimeter and the UBF was replaced by a second etalon in 2000. 16 A fundamental renewal of the G¨ ottingen FPI during the first half of 2005 was the starting point of the development of a new FPI for the 1.5-meter GREGOR solar telescope. 17 New narrow-band etalons and new large-format, high-cadence CCD detectors were integrated into the instrument, accompanied by powerful computer hard- and software. From 2006 to 2007, the optical design for the GREGOR Fabry-P´ erot Interferometer (GFPI) was developed, the necessary optical elements were purchased, and the opto-mechanical mounts were manufactured. 18 An upgrade to full-Stokes spectropolarimetry followed
Transcript
arX
iv:1
207.
2084
v1 [
astr
o-ph
.IM]
9 Ju
l 201
2
GREGOR Fabry-Perot Interferometer – status report and prospects
Klaus G. Puschmanna, Horst Balthasara, Christian Beckb, Rohan E. Louisa, Emil Popowa,Thomas Seelemannc, Reiner Volkmerd, Manfred Wochea, and Carsten Denkera
aLeibniz-Institut fur Astrophysik Potsdam, An der Sternwarte 16, 14882 Potsdam, GermanybInstituto de Astrofısica de Canarias, c/ Vıa Lactea s/n,38205 La Laguna, Spain
The GREGOR Fabry-Perot Interferometer (GFPI) is one of three first-light instruments of the German 1.5-meter GREGORsolar telescope at the Observatorio del Teide, Tenerife, Spain. The GFPI allows fast narrow-band imaging and post-factumimage restoration. The retrieved physical parameters willbe a fundamental building block for understanding the dynamicSun and its magnetic field at spatial scales down to 50 km on thesolar surface. The GFPI is a tunable dual-etalon systemin a collimated mounting. It is designed for spectropolarimetric observations over the wavelength range from 530–860 nmwith a theoretical spectral resolution ofR ≈ 250,000. The GFPI is equipped with a full-Stokes polarimeter. Large-format,high-cadence CCD detectors with powerful computer hard- and software enable the scanning of spectral lines in time spansequivalent to the evolution time of solar features. The field-of-view of 50′′×38′′ covers a significant fraction of the typicalarea of active regions. We present the main characteristicsof the GFPI including advanced and automated calibrationand observing procedures. We discuss improvements in the optical design of the instrument and show first observationalresults. Finally, we lay out first concrete ideas for the integration of a second FPI, the Blue Imaging Solar Spectrometer,which will explore the blue spectral region below 530 nm.
Keywords: Sun — spectroscopy — polarimetry — high angular resolution —instrumentation — image restoration
1. INTRODUCTION
Solar physics has made tremendous progress during recent years thanks to numerical simulations and high-resolution,spectropolarimetric observations with modern solar telescopes such as the Swedish Solar Telescope,1 the Solar OpticalTelescope on board the Japanese HINODE satellite,2 and the stratospheric Sunrise telescope.3 Taking the nature of sunspotsas an example, many important new observational results have been found, e.g., details about the brightness of penumbralfilaments, the Evershed flow, the dark-cored penumbral filaments, the net circular polarization, and the moving magneticfeatures in the sunspot moat. Telescopes with apertures of about 1.5 m such as the GREGOR solar telescope4–7 or theNew Solar Telescope8,9 will help to discriminate among competing sunspot models and to explain the energy balance ofsunspots. New results on the emergence, evolution, and disappearance of magnetic flux at smallest scales can also beexpected. However, these 1.5-meter telescopes are just theprecursors of the next-generation solar telescopes, i.e.,theAdvanced Technology Solar Telescope10 and the European Solar Telescope,11 which will finally be able to resolve thefundamental scales of the solar photosphere, namely, the photon mean free path and the pressure scale height.
Fabry-Perot interferometers (FPIs) have certainly gained importance in solar physics during the last decades, becausethey deliver high spatial and spectral resolution, and a growing number of such instruments is in operation at various tele-scopes. Although most of the instruments have been initially designed only for spectroscopy, most of them have now beenupgraded to provide full-Stokes polarimetry.12–14 The Universitats-Sternwarte Gottingen developed an imaging spectrom-eter for the German Vacuum Tower Telescope (VTT) in the early1990s. This instrument used a universal birefringent filter(UBF) as an order-sorting filter for a narrow-band FPI mounted in the collimated light beam.15 The spectrometer was laterequipped with a Stokes-V polarimeter and the UBF was replaced by a second etalon in 2000.16
A fundamental renewal of the Gottingen FPI during the first half of 2005 was the starting point of the development of anew FPI for the 1.5-meter GREGOR solar telescope.17 New narrow-band etalons and new large-format, high-cadence CCDdetectors were integrated into the instrument, accompanied by powerful computer hard- and software. From 2006 to 2007,the optical design for the GREGOR Fabry-Perot Interferometer (GFPI) was developed, the necessary optical elements werepurchased, and the opto-mechanical mounts were manufactured.18 An upgrade to full-Stokes spectropolarimetry followed
in 2008.19–21 In 2009, the Leibniz-Institut fur Astrophysik Potsdam took over the scientific responsibility for the GFPI,and the instrument was finally installed at the GREGOR solar telescope.22 During the commissioning phase in 2011,three computer-controlled translation stages (two filter sliders and one mirror stage) were integrated and the software wasprepared for TCP/IP communication with external devices according to the Device Communication Protocol (DCP).23 Thispermits automated observing and calibration procedures and facilitates easy operations during observing runs.
2. GREGOR SOLAR TELESCOPE
The 1.5-meter GREGOR telescope is the largest European solar telescope and is designed for high-precision measurementsof dynamic photospheric and chromospheric structures and their magnetic field. Some key scientific topics of the GREGORtelescope are: (1) the interaction between convection and magnetic fields in the photosphere, (2) the dynamics of sunspotsand pores and their temporal evolution, (3) the solar magnetism and its role in solar variability, and (4) the enigmaticheating mechanism of the chromosphere. The inclusion of a spectrograph for stellar activity studies and the search forsolar twins expand the scientific usage of the GREGOR telescope to the nighttime domain.24
The GREGOR telescope replaced the 45-cm Gregory-Coude Telescope, which had been operated on Tenerife since1985. The construction of the new telescope was carried out by a consortium of several German institutes, namely, theKiepenheuer-Institut fur Sonnenphysik in Freiburg, the Leibniz-Institut fur Astrophysik Potsdam (AIP), and the Institutfur Astrophysik Gottingen. In 2009, the Max-Planck-Institut fur Sonnensystemforschung in Katlenburg-Lindau took overthe latter contingent. The consortium maintains international partnerships with the Instituto de Astrofısica de Canarias inSpain and the Astronomical Institute of the Academy of Sciences of the Czech Republic in Ondrejov.
The GREGOR telescope is an alt-azimuthally mounted telescope with an open structure and an actively cooled light-weighted Zerodur primary mirror. The completely retractable dome allows wind flushing through the telescope to facilitatecooling of telescope structure and optics.25 The water-cooled field stop at the primary focus provides a field-of-view (FOV)with a diameter of 150′′. The light is reflected via two elliptical mirrors, and several flat mirrors along the evacuated coudetrain into the optical laboratory. Passing the adaptive optics (AO) system,26 the light is finally distributed to the scientificinstruments. The removable GREGOR Polarimetric Unit (GPU)27 is located near the secondary focus and ensures high-precision polarimetric observations in the visible and near infrared. In the future, the image rotation introduced by thealt-azimuth mount will be compensated by a removable image derotator just behind the exit window of the coude path.A high-order AO system with 196 actuators was recently installed. The closed-loop bandwidth of the system at 0 dB is130 Hz. The AO system uses a Shack-Hartmann wavefront sensorwith 156 sub-apertures. A multi-conjugate AO systemwill be integrated in the near future.26
Three first-light instruments have been commissioned in 2011/12: the GRating Infrared Spectrograph (GRIS),28 theBroad-Band Imager (BBI),7 and the GFPI. A folding mirror deflects the beam either to BBI or to GRIS/GFPI. The lattertwo instruments can be used simultaneously, similar to the multi-instrument setups at the VTT.29 A dichroic beamsplitterdirects wavelengths above 650 nm to the spectrograph, whereas all shorter wavelengths are reflected towards to the GFPI.This beamsplitter can be exchanged with a different beamsplitter with a cutoff above 800 nm. GFPI and BBI are located inan optical laboratory on the 5th floor of the telescope building. GRIS is situated one floor below and receives the light by asecond folding mirror, which is placed just behind the slit unit and the slit-jaw imaging system in the optical laboratory.
3. GREGOR FABRY-PEROT INTERFEROMETER
3.1 Optical design
The GFPI is mounted on five optical tables and is protected by an aluminum housing to prevent the pollution of the opticsby dust and to reduce stray-light. The optical layout of the instrument is shown in Fig.1. Behind the science focusF4, fourachromatic lensesTL1, TL2, HL1, & HL2 create two more fociF5 & F6 and two pupil imagesP1 & P2 in the narrow-bandchannelNBC. The etalonsFPI1 & FPI2 are placed in the vicinity of the secondary pupil in the collimated light beam. Aneutral density filterNDF between the two etalons with a transmission of 63% reduces the inter-etalon reflexes. The beamin theNBC is folded twice byM2 & M5 at a distance of 500 mm and 400 mm fromF5 andHL2, respectively, to minimizethe instrument envelope. Two field stopsFS1 & FS2 reduce stray-light, where the secondary field stop especially avoidsan overlap of the two images created by the removable dual-beam full-Stokes polarimeter. A beamsplitter cubeBS2 nearF5 directs 5% of the light to the broad-band channelBBC. There, the two achromatic lensesTL3 & TL4 are chosen suchthat the image scale of the detectorsCCD1 & CCD2 is exactly the same. A dichroic beamsplitter cubeBS1 just behind the
science focusF4 sends the blue part of the spectrum (below 530 nm) to the imaging channel. One-to-one imaging with thelensesTL1a & TL2a provides the option of recording broad-band images in parallel to GFPI and GRIS observations withtwo pco.4000 camerasCCD1a & CCD2a behind beamsplitterBS2a. This imaging channel will be replaced by the BlueImaging Solar Spectrometer (BLISS) in the future.
Three computer-controlled precision translation stages facilitate automated observing sequences. The two stagesFSL1& FSL2 are used to switch between two sets of interference filters. The filters restrict the bandpass forBBC andNBC to afull-width at half-maximum (FWHM) of 10 nm and 0.3–0.8 nm, respectively. The pre-filters of each channel can be tiltedto optimize the wavelength of the transmission maximum. A third stage inserts a deflection mirrorM1 into the light pathto take calibration data with a continuum light source for spectral calibration purposes. A laser/photo-multiplier channelfor finesse adjustment of the etalons completes the optical setup.
3.2 Cameras, etalons, and control softwareThe GFPI data acquisition system consists of two Imager QE CCD cameras with Sony ICX285AL detectors, which havea full-well capacity of 18,000 e− and a read-out noise of 4.5 e−. The detectors have a spectral response from 320–900 nmwith a maximum quantum efficiency of∼60% at 550 nm. The chips have 1376×1040 pixels with a size of 6.45µm ×
6.45µm. The total chip size is 8.8 mm× 6.7 mm. The image scale at both cameras is 0.0361′′ pixel−1, which leads toa FOV of 49.7′′×37.6′′ in the spectroscopic mode. The maximal blueshift due to the collimated mounting of the etalonsis about 4.32 pm at 630 nm. The cameras are triggered by a programmable timing unit sending out analog TTL signals.The analog-digital conversion is carried out with 12-bit resolution. The data recorded by the cameras are passed via digitalcoaxial cables to the GFPI control computer and are stored ona RAID 0 system.
Two pco.4000 cameras in stock at the observatory can be used for imaging in the blue channel of the FPI. The pco.4000camera has a full-well capacity of 60,000 e− and a read-out noise of 11 e−. The detector has a spectral response from 320–900 nm with a quantum efficiency of∼32% (∼45%) at 380 (530) nm. The pixel size is 9µm × 9 µm, i.e., 4008×2672pixels yield a total chip size of 36 mm× 24 mm. The image scale at both cameras is 0.0315′′ pixel−1. To avoid vignettingof the beam byBS1 andBS2a only 2800×2200 pixels can be used resulting in a FOV of 88.2′′×69.3′′. Three interferencefilters for CaII H λ396.8 nm, Fraunhofer G-bandλ430.7 nm, and blue continuumλ450.6 nm with a FWHM= 1 nm anda transmission better than 60% are available for observations.
The two GFPI etalons manufactured by IC Optical Systems (ICOS) have a diameter of⊘= 70 mm, a measured finesseF ∼ 46, spacingsd = 1.1 and 1.4 mm, and a high-reflectivity coating (R∼ 95%) in the wavelength range from 530–860 nm. The resulting narrow transmission of the instrumentis on the order of FWHM= 1.9–5.6 pm and leads to atheoretical spectral resolution ofR = 250,000. All etalons are operated by three-channel CS100 controllers manufacturedby ICOS. The cavity spacings are digitally controlled by theGFPI control computer via RS-232 communication. Athermally insulated box protects the pupil and the etalons from stray-light and air flows inside the instrument.
The communication between internal (cameras, etalons, andfilter and mirror sliders) and peripheral devices (telescope,AO system, AO filter wheel, GPU, GRIS, etc.) is controlled by the software package DaVis from LaVision in Gottingen,which has been adapted to the needs of the spectrometer.17,23 The modification of the software for TCP/IP communicationwith external devices using DCP allows an easy implementation of automated observing procedures. All observing modessuch as etalon adjustment, line finding, flat-fielding, recording of dark, pinhole, and target images, continuum scans, andrecording of scientific data are now automated.
3.3 Polarimetry at GREGOR
The science verification time in 2012 is mostly devoted to spectroscopy at GREGOR. Polarimetric observations will followin 2013. Nevertheless, a polarization model for the GREGOR telescope has already been developed,21 which is time-dependent because of the alt-azimuthal mount of the telescope. The GPU27 was developed by and built at AIP and canbe inserted at the secondary focus of the telescope to determine the instrumental polarization, which is important for thecalibration of polarimetric measurements. Since 2008, TheGFPI is equipped with a full-Stokes polarimeter19 that can beinserted in front of the detector in theNBC. The polarimeter consists of two ferro-electric liquid crystal retarders (FLCRs)and a modified Savart-plate. The first liquid crystal acts as ahalf-wave plate and the second one as a quarter-wave plate ata wavelength of 630 nm. The modified Savart-plate consists oftwo polarizing beamsplitters and an additional half-waveplate, which exchanges the ordinary and the extraordinary beam. With this configuration, the separation of the two beamsis optimized and the orientation of the astigmatism in both beams is the same so that it can be corrected by a cylindricallens. The present set of FLCRs yields a good efficiency in the spectral range from 580–660 nm. The integration of anautomated calibration procedure for the GFPI will be an important milestone before starting polarimetric observations.
4. GFPI SCIENCE VERIFICATION
For a first characterization of the GFPI performance, we tookseveral data sets in a technical campaign from 15 May to1 June 2012. The AO system was not available because of technical problems with the control computer. Thus, our effortshave been restricted to an optimization of the system and to observations of test data for an estimate of intensity levels,image quality, spectral resolution, stray-light, and other performance indicators of the GFPI and its extended blue imagingchannel. Several problems related to the RS-232 communication with the FPI controllers were resolved so that a stablefinesse is now achieved for several days. In addition, the timing between cameras and etalons has been optimized to ensurethat images are only taken when the etalon spacing has settled to its nominal value.
NBC (1×1 binning) 543.3 40 2400 x x – x x x xBBC (1×1 binning) 543.3 40 – x x – x x x –NBC (2×2 binning) 543.3 30 3500 – – – x x – xNBC (2×2 binning) 557.6 40 2100 – x – x x – xNBC (2×2 binning) 617.3 10 3200 – x – x x – x
CaII H 396.8 15 9230 x x x x x x –G-band 430.7 6 10350 x x x x x x –Blue continuum 450.6 3 11300 x x x x x x –
Note. — Central wavelengthλ0, exposure time∆t, and mean intensityI for all observations (PHA: pinhole array,PH:pinhole,TG: target,QS: quiet Sun,SP: sunspot, andPF: pre-filter).
Table 2. NBC pre-filter characteristics.
Filter λ0 [nm] FWHM [nm] T [%] Binning ∆T [ms] I [counts] Frame rate [Hz]
Note. — Continuum intensitiesI at wavelengthsλ0 for filters with peak transmissionsT and exposure times∆T.
4.1 Imaging spectrometric data
Three complete data sets with 2×2 binning (∼0.0722′′ pixel−1) were taken on 31 May and 1 June, which included imagesof the targetTG and pinholePH mounted in the AO filter wheel atF3 (telescope focal plane). The observations covered thespectral lines at 543.4 nm, 557.6 nm, and 617.3 nm. The FeI λ543.4 nm line had already been scanned on 27 May at fullspatial resolution (0.0361′′ pixel−1) including images of a pinhole array inF4 (science focus at the entrance of the GFPI).SimultaneousBBC images were recorded (Tab.1). Line scans with the GFPI were usually carried out using a step width ofeight digital-analog (DA) steps, whereas the pre-filter scans were performed with one DA step. One DA step correspondsto 0.26–0.41 pm in the spectral range from 530–860 nm.
Broad-band data were taken in the blue imaging channel on 26 and 27 May 2012 just before the GFPI measurements.The observing scheme was the same for all available pre-filters, i.e., 396.8 nm, 430.7 nm, and 450.6 nm. In addition toa few observations of the quiet Sun and a sunspot, images of the target and pinhole inF3, the pinhole-array inF4, andthe pinhole mounted in the GPU atF2 are also included. All data in the blue channel were taken with a combination oftwo spare lenses withf = 500 mm and 1250 mm because a second lens withf = 500 mm was not available, yet. Thus,the image scale of 0.0124′′ pixel−1 oversamples the diffraction limit at these wavelengths by afactor of about three. Theproper image scale of 0.0315′′ pixel−1 will be obtained by one-to-one imaging.
4.2 Intensity estimates for the narrow-band channel
The wavelengthλ0, FWHM, and transmissionT of differentNBC filters are summarized in Tab.2 together with the selectedbinning, the counts at continuum wavelengths, and frame rates at a given exposure time∆T. The results reveal that at fullspatial resolution very long exposure times of up to 100 ms are necessary to achieve at least 2000 counts in the continuumof most of the measured spectral lines for most of the filters with T ∼ 40% and a FWHM∼ 0.3–0.4 nm. As a consequence,one can reach only very low frame rates. A 2× 2-pixel binning speeds up the observations and reduces the exposuretimes. The situation changes when choosing filters with higher transmission. The 617.3 nm filter withT ∼ 80% and aFWHM ∼ 0.74 nm yields reasonable frame rates and exposure times even without binning.
−3 −2 −1 0 1 2 3arcsec
0.0001
0.0010
0.0100
0.1000
1.0000
I/Im
ax
0 5 10 15 20 25arcsec
0.2
0.4
0.6
0.8
1.0
1.2
I/Ic
Figure 2. Derivation of the PSF estimate at 430 nm for the pinhole in F3 (left) and for the complete optical train from a sunspotobservation (right), where the PSF estimates of the four different focal planesare displayed in the lower left corner.
4.3 Estimates of the spatial point spread function
Knowledge of the instrumental point spread function (PSF) provides an estimate of both the spatial resolution and thespatial stray-light level expected for an instrument or telescope.30,31 Using a reference such as a pinhole or a blockingedge in the focal plane, the PSF of all optics downstream can be derived. The combined PSF of the telescope, the post-focus instruments, and the time-variable seeing can also bederived from the observations of a sunspot with its steep spatialintensity gradients.
The PSF of the optical train at the GREGOR telescope relevantfor the GFPI and its complementary imaging channelswas calculated based on reduced images of pinholes located in the focal planesF4, F3, F2, and on sunspot images (seeTab.1). All available calibration images, e.g., target or pinhole images, were averaged for each wavelength step or imageburst. However, only a single image was selected for solar observations because the seeing varies during the imagesequences. The pinhole images were normalized to the maximum intensityImax inside the FOV near the center of thepinhole, whereas the sunspot data were normalized to unity in a quiet Sun region outside the spot. All data were takenwithout real-time correction of the AO system and consequently correspond to the static performance of the optical system.
4.4 Derivation of the point spread function estimates
We took cuts along thex- andy-axes of the CCD across the center of the pinhole in each pinhole observation to obtainan estimate of the PSF. We defined a step function that is assumed to represent the physical extent of the true pinhole.The borders of the step function were set to intersect the observed intensity along the cuts at about the 50% level (leftpanel of Fig.2). We constructed a convolution kernel from a combination ofa Gaussian (varianceσ ) and a Lorentzianfunction (parametera) and convolved the step function with the kernel. A modification of the parameters (σ ,a) withina specific range yielded finally the kernel that best matched the convolved step function and the observed intensity alongthe horizontal and vertical cuts. The cuts inx andy differed far away from the pinhole because of the read-out directionof the CCD camera. Thus, we always tried to find a good match to both thex- andy-cuts close to the pinhole. However,we concentrated only on matching the cut perpendicular to the CCD read-out direction away from the pinhole. Thesame method was applied to all pinhole observations (F4, F3, andF2) to derive a PSF estimate for each focal plane andwavelength. The sunspot observations were modeled by a similar step function at the two transitions from quiet Sun topenumbra and penumbra to umbra (right panel of Fig.2).
The resulting PSF estimates of the four focal planes for the imaging data at 430.7 nm are displayed in the lower leftcorner of the right-hand panel of Fig.2. The width of the PSF estimates is similar forF4 andF3, where only static opticalcomponents and negligible seeing effects contribute to thePSF and a diffraction-limited performance can be reached, butincreases significantly when passing to the focal planeF2 that experiences telescope seeing and seeing fluctuations alongthe coude train. The PSF derived from the sunspot observations includes all seeing effects and roughly doubles its widthrelative to the PSF atF2.
Figure 3. Energy enclosed within the radiusr for the PSF estimates at 430.7 nm of the four different focal planes (left). Thedashedvertical linesdenote the radius where 90% of the energy is enclosed. Thedotted vertical linedenotes the diffraction limit at 430.7 nm.Similar observations were carried out at different wavelength (right). Theoretical values are given in parentheses.
By applying the approach described above to all available data, PSF estimates for all four focal planes were obtainedin case of the imaging data at 430.7 nm, and for some of the focal planes in case of the GFPIBBC andNBC data (see tablein Fig. 3). The optical performance and the expected stray-light level can best be quantified from the total energy enclosedin a given radius, i.e., a radial integration of the PSF whosevolume yields the amount of light spread up to a given radius.Figure3 shows the enclosed energyE for the PSF estimates at 430.7 nm. From the curves, we derivedtwo values, namely,the energy enclosed at the radiusE(rDL) of the diffraction limit and the radiusr(E = 90%) at which 90% of the energy isenclosed. The former value provides an estimate of the generic stray-light to be expected by subtracting it from 100%. Thelatter value can be used as generic estimate of the spatial resolution. The table in Fig.3 lists the two values for all analyzedwavelengths and channels, i.e., the blue imaging channel, and the GFPIBBC andNBC at specific wavelengths.
A comparison between the value of the diffraction limit and the radius where 90% of the energy is enclosed shows thatthe optics behindF4 performs close to diffraction limit. The optics downstreamof F3 performs slightly worse with a totalenclosed energy of about 70–75% of the diffraction-limitedcase. This implies a generic spatial stray-light level – if onedefines stray-light as all light scattered to outside a distance of one times the diffraction limit – of about 25% created bythe optics downstream ofF3. The corresponding value atF4 is about 10–15%. The values of both the stray-light leveland the radius that encloses 90% of the total energy experience a profound jump when passing toF2 and beyond. Allthese data were taken at mediocre seeing conditions and without AO correction. The latter will be necessary for a finalcharacterization of the optical performance of the complete optical train including the telescope.
4.5 Spectral resolution, spectral stray-light, and blue-shift
The spectral resolution and thespectralstray-light inside theNBC was estimated by a convolution of Fourier Trans-form Spectrograph (FTS) atlas spectra with a Gaussian of width σ and a subsequent addition of a constant wavelength-independent stray-light offsetβ .32,33 This component of stray-light corresponds to light scattered onto the CCD detectorwithout being spectrally resolved. Therefore, it changes the line depth of observed spectral lines. The convolved FTS spec-tra in each wavelength range were compared with two sets of spatially averaged observational profiles that either coveredthe full pre-filter transmission curve or only the line inside the same range that is usually recorded in science observations(543.4 nm, 557.6 nm, 617.3 nm). The left panel of Fig.4 shows the average observed spectrum at FeI λ617.33 nm, theoriginal FTS spectrum, and the FTS spectrum after the convolution with the best-fit Gaussian kernel and the addition of thestray-light offset. The method has some ambiguity betweenσ andβ , which can be modified in opposite directions oversome range near the best-fit values without significantly degrading the reproduction of the observed spectra. Therefore, thevalues listed in Tab.3 have an error of about±0.5 pm inσ and±5% inβ . The stray-light levelβ inside the GFPI is foundto be below 10% and the spectral resolution isR ∼ 100,000, quite below the theoretically expected value ofR ∼ 250,000.The dispersion values derived from the observed spectra arelisted in Tab.3 and correspond to eight DA steps. They matchthe theoretically expected values given in parentheses.
Table 3. Spectral characteristics of the GFPI data.
Figure 4. Close-up of the FeI λ617.3 nm spectral line at reduced spectral sampling (left): observed spectrum (black pluses), FTS atlasspectrum (red solid line), FTS convolved with the theoretical Airy function (orange dash-dotted line), and FTS convolved with the best-fit Gaussian kernel (blue dashed line). An appropriate stray-light offset (β ′,β ) was added after the convolution. The Airy function andthe Gaussian kernel in the left corner are displayed in arbitrary units. Scan of the pre-filter curve around theλ543.4 nm line at maximalspectral sampling (right).
For comparison, the FTS spectrum was convolved with the theoretical Airy function that describes the transmissionthrough the double-etalon system using the measured finesseof 46. In this case, the line depth after the convolution could beadjusted by a stray-light offsetβ ′ of about 20%, but the observed line width in the wings of the line is not matched (Fig.4).The performance of the GFPI in terms of the spectral resolution depends on the exact alignment of the plate parallelism ofthe etalons. For any detailed modeling of the spectral transmission it seems recommended to use a comparison of observedand atlas spectra instead of relying on the theoretically expected spectral PSF.
The measured finesse was obtained with a laser light source that produces a beam of about 15 mm diameter passingthrough the etalons. Therefore, the parallelism of the etalon plates was only optimized in the central part. In case of theobserved spectra, the actual pupil diameter ofP2 in the GFPI is about 60 mm and samples nearly the full surface of theetalons, in contrary to the VTT setup, where the pupil in the vicinity of the etalons still had a diameter of only about 40 mm.In the latter case, a spectral resolution still ofR ∼ 160,000 was obtained applying the same method (see Tab.3).
A degradation of the finesse between the central and peripheral regions of about 1.3 was found for a single etalon forradii of 15 mm and 50 mm.34 The spectral resolution in the central area of the etalons ofthe GFPI at GREGOR wouldtherefore beR ∼ 170,000, which is clearly above the thermal line width (λ/σth ≈ 134,000). Note, however, that in thisestimate similar effects of the second etalon are not yet considered. Thus, the effective spectral resolution of the centralarea might be even close to the value expected from theory. A widening of the laser beam, and consequently the adjustmentof the finesse over a broader area fraction, might significantly improve the results.
The maximal blueshift induced in the spectra because of placing the FPIs inside a collimated beam was determinedfrom a set of flat-field data. The numbers are close to the theoretically expected values (last column of Tab.3).
Table 4. Selected spectral lines in the range 380–500 nm.
Ion Wavelengthλ Excitation potentialχ Lande-factorg Equivalent width Comment
FeI 384.998 nm 1.01 eV 0.00 60.8 pm blendsCN-band 388.300 nmHe 388.865 nm 19.73 eV only prominencesHζ 388.905 nm 10.20 eV 234.6 pm only prominencesFeI 392.922 nm 3.25 eV 2.50 3.7 pm weak lineCaII K 393.368 nm 0.00 eV 2025.3 pmCaII H 396.849 nm 0.00 eV 1546.7 pmHε 397.008 nm 10.20 eV line wing CaII HFeI 404.583 nm 1.48 eV 117.5 pm blendsFeI 406.539 nm 3.43 eV 0.00 6.4 pmMn I 407.028 nm 2.19 eV 3.33 6.6 pm high prioritySr II 407.772 nm 0.00 eV 42.8 pm resonance lineFeI 408.088 nm 3.29 eV 3.00 6.1 pm high priorityHδ 410.175 nm 10.20 eV 313.3 pmFeII 430.318 nm 2.70 eV 1.47 10.3pm in G-bandG-band 430.500 nmHγ 434.048 nm 10.20 eV 285.5 pmFeI 440.476 nm 1.56 eV 89.8 pm low priorityBa II 455.404 nm 0.00 eV 15.9 pm resonance lineMg I 457.110 nm 0.00 eV 9.2 pm resonance lineFeI 461.321 nm 3.29 eV 0.00 6.6 pm neargeff = 2.5 lineCr I 461.367 nm 0.96eV 2.50 6.2 pm neargeff = 0 lineTi I 464.519 nm 1.73 eV 2.50 1.6 pm enhanced in spotsFeI 470.495 nm 3.69 eV 2.50 5.8 pm high priorityHβ 486.134 nm 10.20 eV 368.0 pmFeI 470.178 nm 3.93 eV 1.50 4.4 pm neargeff = 0 lineNi I 491.203 nm 3.77 eV 0.00 4.7 pm
The scans of the pre-filter transmission curve at maximal spectral sampling showed another spectral feature in the GFPIdata whose origin remains unclear up to now. A beat with stable period and varying amplitude is superimposed on thespectral lines (right panel of Fig.4 for 543.4 nm). This beat is observed in all spectra, independent of wavelength and pre-filter, and has been present since the integration of the second narrow-band etalon in 2007. Owing to the aforementionedmodifications, the beat is now absolutely stable with time inposition and amplitude for each filter. Forward or backwardscanning of the spectrum yields identical results. Thus, a removal of the beat is straightforward using the white-lightsourceas a reference.
5. BLUE IMAGING SOLAR SPECTROMETER
The spatial resolution of a telescope scales inversely proportional with the observed wavelength. Therefore, observationsat short wavelengths (below 530 nm) offer the opportunity toobtain data with higher spatial resolution. This advantageis partly diminished by the seeing degradations and the smaller number of photons at shorter wavelengths. There arerelatively few ground-based instruments for spectral observations in the blue spectral region.35–38 All of these observationswere carried out with slit-spectrographs that permit only limited improvements by post-factum restoration techniques.30 AFabry-Perot-based imaging spectrometer provides data suitable for sophisticated post-factum image restoration techniquesto yield high spatial and spectral resolution data. This is the motivation for BLISS, which will supplant the blue imagingchannel of the GFPI in the near future.
5.1 Interesting spectral lines in the blue part of the visible spectrum
In the wavelength range covered by BLISS, there are several spectral lines and two molecular bands of high scientificinterest (Tab.4 and Fig.5). All Balmer-lines except Hα are at shorter wavelengths than 530 nm, and the H and K linesof ionized calcium are the strongest lines in the visible part of the solar spectrum probing the chromosphere. Severalphotospheric resonance lines such as MgI λ457 nm, SrII λ407 nm, and BaII λ455 nm are found in blue part of the visiblespectrum. Here, one also finds several magnetically insensitive lines (geff = 0)39 and on the other hand lines exhibiting aZeeman-triplet with splitting factors ofgeff = 2.5 and higher.40 A special case is the pair FeI λ461.32 nm (geff = 0) andCr I λ461.37 nm (geff = 2.5) that can be recorded at the same time. The molecular bands of CH λ430 nm (G-band) andCN λ388 nm are well suited to investigate very small hot featuresin the solar photosphere.
0.0
0.2
0.4
0.6
0.8
1.0
1.2
Inte
nsity
I / I
0
Fe I He Fe I Fe I Fe I Mn I Sr II Fe I Fe II Fe I Ba II Mg I Fe I Ti I Fe I Ni I385.00 388.86 392.92 404.58 406.54 407.03 407.77 408.09 430.32 440.48 455.40 457.11 461.32 464.52 470.49 491.20
0.0
0.2
0.4
0.6
0.8
1.0
1.2
Inte
nsity
I / I
0
Hζ Hε Hδ Hγ Hβ388.90 397.01 410.17 434.05 486.13
0.0
0.2
0.4
0.6
0.8
1.0
1.2
Inte
nsity
I / I
0
CN−Band388.30
G−Band430.50
0.0
0.2
0.4
0.6
0.8
1.0
1.2
Inte
nsity
I / I
0
Wavelength λ
Ca II K393.37
Ca II H396.85
Figure 5. Selected spectral lines in the range 380–500 nm (see Tab.4). The name and central wavelength of all spectral lines are given atthe top of each panel. Major (minor) tick marks are separatedby 0.1 nm (10 pm). Note that the spectral coverage of the panels changesfrom row to row, i.e., 0.1 nm, 0.2 nm, 1.0 nm, and 1.0 nm (from top to bottom).
5.2 Optical design
The geometrical design of BLISS is depicted in Fig.6, which shows its integration into the GFPI by replacing the blueimaging channel. Most of the optical components and their labels are identical to those in Fig.1 so that any informationcan be directly taken from there. The calculations for this initial optical design revealed that a pupilP2 of about 70 mmdiameter will be sufficient for an acceptable maximal blue-shift over the entire wavelength range at a given image scale andFOV. Thus, the design of BLISS itself is almost identical to the design of the GFPI, apart from a re-design of the cameralensesHL2 andTL4 in the NBC andBBC to obtain an adequate image scale. The schematical optical designs of bothinstruments are compared in Fig.7. The diameter of the light beam at the different foci, pupils, and on the relevant opticalsurfaces has been calculated by means of geometrical opticsand confirmed by ZEMAX ray-tracing as for the GFPI.18 TheZEMAX ray-tracing is presented in Fig.8.
If one considers the actual values of the usable diameter of the main mirror and the focal length of the GREGORtelescope, the size of the pupilP2 inside the GFPI (59 mm) is smaller than the value assumed in the original calculation in2007 (63 mm). However, a modification of the parabolic collimator and camera mirrors of the AO system will bring thepupil diameter again closer to the original value. Nevertheless, by changing the focal length ofTL2, the reduction of thepupil diameter is currently compensated in the design of BLISS (see Fig.7).
Table 1
Table 2
Table 3
Table 4 Table 5
F4
FS1
BS1TL1
TL1a
P1
P1a
TL2
TL2a
F5/FS2
F5a/FS2a
FSL1
FSL1a
M2
M2a
HL1
HL1a
FPI1
FPI1a
NDF
NDFa
P2
P2a
FPI2
FPI2a
HL2
HL2aM5
M5a
POL
CCD1/F6
POLa
CCD1a/F6a
BS2
BS2a
TL3
TL3a
TL4
TL4a
FSL2
FSL2a
CCD2/F6
CCD2a/F6a
M4MP
M3
LFS
LPF
Laser LL1
LL2
M4a
MPa
M3a
LFSa
LPFa
Laser_a
LL1a
LL2a
Figure 6. Geometrical design for the integration of BLISS into the GFPI system. The design is based on the study presentedin Fig. 7.The optical elements of BLISS are labeled with an extra “a”. Apart fromTL2a, HL2a, andTL4a, they are identical to those of the GFPI.
A new type of cameras has been envisaged for BLISS, namely, two sCMOS-cameras from PCO (pco.edge). Thesecameras have 2560×2160 pixels with a pixel size of 6.5µm× 6.5µm similar to the GFPI Imager QE cameras. A changeof the focal lengths ofHL2 from f = 1500 mm tof = 2250 mm and ofTL4 from f = 600 mm tof = 900 mm for BLISS(see Fig.7) yields an image scale of 0.0276′′ pixel−1 and a FOV of 70.6′′×59.6′′ on both cameras. With this configuration,the maximal blue-shift across the FOV amounts to 4.44 pm and 6.19 pm at 380 nm and 530 nm, respectively.
In addition, we checked the possibility of interchanging the sCMOS and Imager QE cameras between the GFPI andBLISS. The almost identical pixel sizes of both cameras facilitates this task. For the integration of the sCMOS cameras intothe GFPI, a circular FOV with a diameter ofd= 78.8′′ and an image scale of 0.0364′′ pixel−1 for spectroscopy would resultin a maximal blueshift of 6.91 pm at 630 nm. However, the FOV for polarimetry is limited by the full-Stokes polarimeterand would remain at its previous size of 24.9′′×37.6′′. The frame rate at full resolution would increase from 7 to 40Hz,which perfectly suits this observing mode.
The integration of the Imager QE cameras into BLISS would result in an image scale of 0.0274′′ pixel−1 and a FOVof 37.6′′× 28.5′′ with a maximum blueshift of 1.15 pm and 1.61 pm at 380 nm and 530nm, respectively. The lowerframe rates would match the longer exposure times at shorterwavelength. A super-achromatic optical setup for BLISS –as also foreseen for the GFPI – would be preferable but otherwise all lenses exceptTL4 of BLISS could be purchased asoff-the-shelf achromats.
Figure 7. Schematical optical design (top) of the current GFPINBC, BBC, and blue imaging channel. In the future, the BLISSNBCandBBC (bottom) will replace the GFPI blue imaging channel. The diameter ofthe beam at all optical surfaces, the focal length of thelenses, and the positions of other optical elements are given in millimeter. The achromatic lensesTL2, HL2, andTL4 of BLISS have afocal length of 220 mm, 2250 mm, and 900 mm, respectively. ThemirrorsM2 andM5 of the GFPI will be moved for the integration ofBLISS by 200 mm and 50 mm, respectively.
The distribution of the two instruments on the five optical tables in the GREGOR observing room is shown in Fig.6.All optical elements of BLISS have been labeled with an additional “a” to distinguish them from those of the GFPI. Behindthe common dichroic beamsplitter cubeBS1, each instrument has its ownNBC andBBC. A displacement of the foldingmirrorsM2 andM5 of the GFPI to a distance of 700 mm and 350 mm fromF5 andHL2, respectively, yields some freespace on optical table 3 that can be used for theNBC of BLISS. The ZEMAX ray-tracing revealed changes in the opticalpath when considering the etalons plates in the design. Thus, the GFPINBC has to be shortened further by reducing thedistance betweenP2 andHL2 from 500 mm to 350 mm. This detail is not considered in Figs.6 and7.
The beam in theNBC of BLISS is also folded twice byM2a andM5a at a distance of 800 mm and 700 mm fromF5a andHL2a, respectively. In theBBC of BLISS,TL3a andTL4a are separated by 350 mm, in contrary to the 50 mm betweenTL3andTL4 in case of the GFPI. Two computer-controlled filter slidersFSL1a andFSL2a will again switch between two setsof interference filters, which restrict the bandpass for theBLISSBBC andNBC. BLISS is mainly designed for spectroscopybecause of the expected low photon numbers in the blue spectral region. Nevertheless, a full-Stokes polarimeter can easilybe integrated into the system. As in case of the GFPI, a laser/photo-multiplier channel for finesse adjustment of the etalonswill be implemented. The white-light channel of the GFPI will be removed and be replaced by an external white-lightsource common to all instruments at GREGOR.
Figure 8. Total arrangement of GFPI and BLISS in a ZEMAX multi-configuration file in shaded modeling for the respective centralwave length of each instrument and their maximal field dimension. The design confirms the calculations in geometrical optics presentedin Figs.6 and7.
5.3 Camera system
Modern sCMOS cameras are capable of delivering high frame rates using large-format sensors with low readout noise,which makes them ideally suited for an application in solar physics. The pco.edge is a potential candidate for BLISS.The sensor of this camera has 2560×2160 pixels with a size of 6.5µm × 6.5 µm and a full well capacity of 30,000 e−.The camera has a 16-bit digitization and would be operated ina global shutter mode. In this mode, the camera has amaximum frame rate of 40 Hz. The cameras are running in a “fastscan mode” (286 MHz) because of speed limitationsof the camera link interface. The 16-bit signal is internally converted to 12-bit in the fast scan mode. Inside the interfaceit is decompressed again to 16-bit. The signal losses due to the compression are roughly a factor of ten smaller than theshot noise of the camera signal. The readout noise is in the order of 2.3 e−. The dark current in the global shutter modeconsists of a part related to the exposure time, i.e., 2–6 e− pixel−1 s−1 and a part related to the sensor readout time, which isconstant for a given pixel clock, i.e., 0.6 e− pixel−1 in the fast scan mode. Peltier cooling of the sensor ensures an operatingtemperature of+5◦ C. The camera has a quantum efficiency of∼30% and∼54% at 380 nm and 530 nm, respectively,similar to the Imager QE cameras currently used in the GFPI (∼36% and∼60%).
To handle the extremely large data bandwidth of about 300 MB s−1 at frame rates of 40 Hz, each camera will becontrolled by an individual PC, in which the data will be stored on local RAID 0 systems. The integration of the camerasin the control software is straightforward because the DaVis software will be operational on the new system with minorchanges only. The cameras are currently being tested at AIP including, e.g., an analysis of image quality, power spectra,and noise characteristics.
However, the feasibility of using the pco.edge for BLISS hasstill to be demonstrated. A detailed photon statisticin the wavelength range 380–530 nm will help with the final decision, if these cameras will be employed in BLISS orthe GFPI. For the blue wavelength range, rather long exposure times can be expected, making the use of high-speedcameras somewhat doubtful. Moreover, the exposure time of the cameras currently has an upper limit of 100 ms due to arelatively high dark current in the global shutter mode. On the other hand, high frame rates would be extremely beneficialwhen operating the GFPI in the vector polarimetric mode, because at present this instrument is limited to just 5–7 Hzat full resolution. The disadvantage would be just a partialusage of the chip. The almost identical pixel size facilitatesinterchanging the cameras between the two instruments.
599.0 599.5 600.0 600.5 601.0Wavelength λ [nm]
10−8
10−6
10−4
10−2
100
Inte
nsity
I / I
0
399.0 399.5 400.0 400.5 401.0Wavelength λ [nm]
10−8
10−6
10−4
10−2
100
Inte
nsity
I / I
0
Figure 9. Comparison of the GFPI (left)/BLISS (right) transmission profiles. The curves correspond to a narrow-band interference filter(FWHM = 0.3 nm, thin solid line), etalon 1 (R= 0.95/0.92, d = 1.4/0.41 mm, FWHM= 2.1/5.2 pm, thin dashed line), etalon 2(R= 0.95/0.92, d = 1.1/0.24 mm, FWHM= 2.7/8.9 pm, thin dash-dotted line), and all transmission curves combined (FWHM=1.5/4.2 pm, parasitic-light fraction 0.8/0.4%,thick solid line). The transmission profiles are normalized to unity at the central wavelengthλ0 = 600.0/400.0 nm.
5.4 Etalons
Airy functions are quasi-periodic, i.e., in a dual etalon system their characteristics parameters (reflectivityR and plateseparationd) have to be carefully chosen to minimize parasitic light from neighboring orders. Even if an optimal choicehas been identified, narrow-band interference filters (0.3–0.8 nm) have to be used to minimize the contributions by theparasitic light. In general, an optimization of the characteristic FPI parameters has to include realistic transmission curvesfor the interference filters. In Fig.9, we present the results of such a parameter study for BLISS and compare themwith the case of the GFPI. The characteristic parameters of the BLISS etalons were chosen such that the parasitic light isminimized. The reflectivity of etalons for imaging in the blue spectral region has to be lower to increase the FWHM andto accommodate the smaller number of available photons. Therefore, the rejection of parasitic light further away from thecentral wavelengthλ0 becomes more important and narrower interference filters are required. The combination of etalonswith d = 0.41 and 0.24 mm (FWHM= 5.2 and 8.9 pm), will further increase the light level and result in a theoreticalspectral resolution ofR ∼ 100,000.
6. DATA PIPELINE FOR IMAGING SPECTROPOLARIMETRY
Analyzing data from imaging spectropolarimetry requires intimate knowledge about seeing, telescope, Fabry-Perot etalons,detector systems, instrumental polarization, and image restoration techniques. This knowledge about the entire imageformation process has to be encapsulated in a reliable and robust data pipeline, which provides the user with well calibrated,self-describing data suitable for further analysis using,for example, spectral line inversion techniques.
The GFPI builds upon a two-decade heritage of Fabry-Perot interferometers operated at the VTT. During this periodthe data reduction code bifurcated significantly necessitating rewriting the code from scratch. The best features of theindividual codes were kept and in addition, the most time-consuming algorithm were tuned for performance.
6.1 Data processing and image restoration
Imaging spectropolarimetry12,23 offers the possibility to improve the recorded data beyond AO real-time correction. There-fore, several state-of-the-art image restoration techniques will be part of the GFPI data pipeline. At present, two of the mostsuccessful solar image restoration tools are included, namely, the Gottingen speckle imaging and deconvolution code41–43
and Multi-Frame Blind Deconvolution44 (MFBD) as well as its extension to multiple objects45 (MOMFBD). As a thirdoption the Kiepenheuer-Institute Speckle InterferometryPackage46,47 (KISIP) will be included in the near future.
The data pipeline was developed in particular for data of theGFPI and BLISS. However, its modular approach to imageprocessing facilitates the integration of other imaging spectropolarimeters. The I/O interface can be easily adapted. Since
image restoration is an integral part of the data pipeline, also images from high-cadence, large-format CCD or CMOScameras can be easily reduced. The information about observatories, telescopes, instruments, and detectors is kept inconfiguration files, which can simply be modified for site-specific needs.
The data pipeline is based on the Interactive Data Language (IDL). We adhere to a slightly modified version of the IDLcoding standard48 to provide the same “touch-and-feel” for all programs. Thisis enforced by using the same variable namesin all programs, where descriptive variable names are the guiding principle rather than short names without meaning. TheIDL source code of the GFPI data pipeline is available to all GFPI users. However, the hardware requirements go beyondthat provided by typical work stations. Nevertheless, scaled-down data processing is still possible on a desk-top computer.
6.2 Data formats and data conversion
GFPI data are written in a native DaVis format, which uses on-the-fly image compression to achieve high data rateswhile writing data to a RAID 0 harddisk array. Narrow- and broad-band images are saved together in one file for eachsimultaneous exposure. Imaging spectropolarimetric sequences are accompanied by “set” files, i.e., an ASCII file with allinformation describing the data and observing modes. Thesecompressed data are archived at AIP, whereas the principleinvestigator (PI) of the observing run receives a copy of thedata in the Flexible Images Transport System49 (FITS) formatusing image extensions.50 Data from the GFPI data acquisition computer can be transferred by FTP to servers for datastorage and processing, which are located at the VTT. The transfer lasts about the same time as taking the data. The rawdata are then converted to FITS. Both data types are stored finally to LTO-Ultrium tapes with a storage capacity of about800 GB.
6.3 Quick-look data, on-line repositories, and automatic logbooks
The data conversion step also includes several other tasks:(1) Image statistics are computed and included in the headersof the image extensions. These values are used in an heuristic error analysis to identify potential problems in the dataacquisition process early on. Together with quantities describing the GFPI performance, they are stored in a database tomonitor the long-term stability of the instrument and the quality of its data products. (2) Quick-look data products canbecomputed in time spans that are comparable to the data acquisition. Some functionality is already provided by the DaVissoftware (visualization of spectral scans, contrast enhancement, or region-of-interest manipulations). However, line-of-sight velocity maps or vector magnetograms have to be computed off-line. Limiting the image restoration to just a simpledestretching of narrow- and broad-band images and taking averages at successive wavelength positions already providesspatially resolved information about the physical properties of the observed solar features. (3) Quick-look data are offeredin two formats, either as web pages or as logbooks in the Portable Document Format (PDF), which both document eachday’s observing run. The web pages with quick-look data willbecome public immediately, whereas the raw and FITS datawill be embargoed for a certain time (typically one year), ifthe data are taken in the PI-mode.
6.4 Observing modes
Data from imaging spectropolarimetry can be analyzed in many different ways, for example, different image restorationmethods are available, noise in both narrow- and broad-bandimages can be treated differently, or different schemes forthe polarimetric correction of the data can be applied. All parameters describing a particular data processing scheme aresaved in “project” files, which also serve as templates for observing runs with similar characteristics. This ensures that dataproducts from different observing runs are comparable and enables their use in databases, where a user expects to havedata products of the same quality.
Post-focus instruments at ground-based solar observatories are often operated in PI-mode, where the PI of an observingrun specifies the instrument setup and observing sequences.The data are considered property of the PI and often only asmall fraction of data finds its way into scientific publications. Changing this type of approach requires a concerted effort,both on the instrument-builder side and on the part of the developers of data pipelines. The GFPI control software offersa user interface for all standard data acquisition modes (dark, flat-field, target, and pinhole frames as well as spectrallinescans). Interaction with the telescope, its subsystems, and other peripheral devices is handled automatically for each task.The “set” files for each mode contain all necessary information for further processing the data. The intermediate step ofconverting the data to FITS format allows the users to use their own data processing routines in addition to the GFPI datapipeline. A detailed account of the data processing is beyond the scope of this conference proceedings and will be deferredto a data release paper in a peer-reviewed journal.
ACKNOWLEDGMENTS
The 1.5-meter GREGOR solar telescope was built by a German consortium under the leadership of the Kiepenheuer-Institutfur Sonnenphysik in Freiburg with the Leibniz-Institut f¨ur Astrophysik Potsdam and the Max-Planck-Institut fur Sonnen-systemforschung in Katlenburg-Lindau as partners, and with contributions by the Instituto de Astrofısica de Canarias, theInstitut fur Astrophysik Gottingen, and the Astronomical Institute of the Academy of Sciences of the Czech Republic.CDwas supported by grant DE 787/3-1 of the Deutsche Forschungsgemeinschaft (DFG).
REFERENCES
[1] Scharmer, G. B., Bjelksjo, K., Korhonen, T. K., Lindberg, B., and Petterson, B., “The 1-meter Swedish solar tele-scope,” in [Innovative telescopes and instrumentation for solar astrophysics], Keil, S. L. and Avakyan, S. V., eds.,Proc. SPIE4853, 341–350 (2003).
[2] Kosugi, T., Matsuzaki, K., Sakao, T., Shimizu, T., Sone,Y., Tachikawa, S., et al., “The Hinode (Solar-B) mission: anoverview,”Sol. Phys.243, 3–17 (2007).
[3] Solanki, S. K., Barthol, P., Danilovic, S., Feller, A., Gandorfer, A., Hirzberger, J., et al., “SUNRISE: instrument,mission, data, and first results,”ApJL723, L127–L133 (2010).
[4] Volkmer, R., von der Luhe, O., Kneer, F., Staude, J., Balthasar, H., Berkefeld, T., et al., “New high resolutionsolar telescope GREGOR,” in [Modern solar facilities – advanced solar science], Kneer, F., Puschmann, K. G. &Wittmann, A. D., ed., 39 (2007).
[5] Volkmer, R., von der Luhe, O., Denker, C., Solanki, S., Balthasar, H., Berkefeld, T., et al., “GREGOR telescope:start of commissioning,” in [Ground-based and airborne telescopes III], Stepp, L. M., Gilmozzi, R., and Hall, H. J.,eds.,Proc. SPIE7733, 77330K (2010).
[6] Volkmer, R., von der Luhe, O., Denker, C., Solanki, S. K., Balthasar, H., Berkefeld, T., et al., “GREGOR solartelescope: design and status,”Astronomische Nachrichten331, 624 (2010).
[7] Schmidt, W., von der Luhe, O., Volkmer, R., Denker, C., Solanki, S., Balthasar, H., et al., “The GREGOR solartelescope on Tenerife,” ASP Conf. Ser., in press,ArXiv e-prints, 2012arXiv1202.4289S(2012).
[8] Denker, C., Goode, P. R., Ren, D., Saadeghvaziri, M. A., Verdoni, A. P., Wang, H., et al., “Progress on the 1.6-meterNew Solar Telescope at Big Bear Solar Observatory,” in [Ground-based and airborne telescopes], Stepp, L. M., ed.,Proc. SPIE6267, 62670A (2006).
[9] Cao, W., Gorceix, N., Coulter, R., Coulter, A., and Goode, P. R., “First light of the 1.6 meter off-axis New SolarTelescope at Big Bear Solar Observatory,” in [Ground-based and airborne telescopes III], Proc. SPIE7733, 773330-773330-8 (2010).
[10] Wagner, J., Rimmele, T. R., Keil, S., Hubbard, R., Hansen, E., Phelps, L., et al., “Advanced Technology SolarTelescope: a progress report,” in [Ground-based and airborne telescopes II], Stepp, L. M. and Gilmozzi, R., eds.,Proc. SPIE7012, 70120I (2008).
[11] Collados, M., Bettonvil, F., Cavaller, L., Ermolli, I., Gelly, B., Grivel-Gelly, C., et al., “European Solar Telescope:project status,” in [Ground-based and airborne telescopes III], Stepp, L. M., R., G., and Hall, H. J., eds.,Proc. SPIE7733, 77330H (2010).
[12] Puschmann, K. G. and Beck, C., “Application of speckle and (multi-object) multi-frame blind deconvolution tech-niques on imaging and imaging spectropolarimetric data,”A&A 533, A21 (2011).
[13] Martınez Pillet, V., Del Toro Iniesta, J. C.,Alvarez-Herrero, A., Domingo, V., Bonet, J. A., Gonzalez Fernandez,L., et al., “The Imaging Magnetograph eXperiment (IMaX) forthe Sunrise balloon-borne solar observatory,”Sol.Phys.268, 57–102 (2011).
[14] Rimmele, T. R., Wagner, J., Keil, S., Elmore, D., Hubbard, R., Hansen, E., et al., “The Advanced Technology SolarTelescope: beginning construction of the world’s largest solar telescope,” in [Ground-based and Aiborne TelescopesIII ], Stepp, L. M. and Gilmozzi, R., and Helen, H., eds.,Proc. SPIE7733, 77330G-17 (2010).
[15] Bendlin, C., Volkmer, R., and Kneer, F., “A new instrument for high resolution, two-dimensional solar spectroscopy,”A&A 257, 817–823 (1992).
[16] Volkmer, R., Kneer, F., and Bendlin, C., “Short-periodwaves in small-scale magnetic flux tubes on the Sun.,”A&A 304, L1 (1995).
[17] Puschmann, K. G., Kneer, F., Seelemann, T., and Wittmann, A. D., “The new Gottingen Fabry-Perot spectrometerfor two-dimensional observations of the Sun,”A&A 451, 1151–1158 (2006).
[18] Puschmann, K. G., Kneer, F., Nicklas, H., and Wittmann,A. D., “From the ”Gottingen” Fabry-Perot interferometerto the GREGOR FPI,” in [Modern solar facilities – advanced solar science], Kneer, F., Puschmann, K. G., andWittmann, A. D., eds., 45 (2007).
[19] Bello Gonzalez, N. and Kneer, F., “Narrow-band full Stokes polarimetry of small structures on the Sun with specklemethods,”A&A 480, 265–275 (2008).
[20] Balthasar, H., Bello Gonzalez, N., Collados, M., Denker, C., Hofmann, A., Kneer, F., and Puschmann, K. G., “Afull-Stokes polarimeter for the GREGOR Fabry-Perot interferometer,” in [Cosmic Magnetic Fields: From Planets,to Stars and Galaxies], Strassmeier, K. G., Kosovichev, A. G., and Beckman, J. E.,eds.,IAU Symp.259, 665–666(2009).
[21] Balthasar, H., Bello Gonzalez, N., Collados, M., Denker, C., Feller, A., Hofmann, A., et al., “Polarimetry withGREGOR,” in [Solar polarization 6], Kuhn, J. R., Harrington, D. M., Lin, H., Berdyugina, S. V.,Trujillo-Bueno, J.,Keil, S. L., and Rimmele, T., eds.,ASP Conf. Ser.437, 351 (2011).
[22] Denker, C., Balthasar, H., Hofmann, A., Bello Gonzalez, N., and Volkmer, R., “The GREGOR Fabry-Perot interfer-ometer: a new instrument for high-resolution solar observations,” in [Ground-based and airborne instrumentation forastronomy III], McLean, I. S., Ramsay, S. K., and Takami, H., eds.,Proc. SPIE7735, 77356M–77356M–12 (2010).
[23] Puschmann, K. G., Balthasar, H., Bauer, S. M., Hahn, T.,Popow, E., Seelemann, T., et al., “The GREGOR Fabry-Perot Interferometer – a new instrument for high-resolution spectropolarimetric solar observations,” ASP Conf. Ser.,in press,ArXiv e-prints, 2011arXiv1111.5509P(2011).
[24] Strassmeier, K. G., Woche, M., Granzer, T., Andersen, M. I., Schmidt, W., and Koubsky, P., “Gregor@Night,” in[Modern solar facilities – advanced solar science], Kneer, F., Puschmann, K. G. & Wittmann, A. D., ed., 51 (2007).
[25] Bettonvil, F. C. M., Hammerschlag, R. H., Jagers, A. P.L., and Sliepen, G., “Large fully retractable telescope enclo-sures still closable in strong wind,” in [Advanced Optical and Mechanical Technologies in Telescopes and Instrumen-tation], Proc. SPIE7018, 70181N-70181N-9 (2008).
[26] Berkefeld, T., Soltau, D., Schmidt, D., and von der Luhe, O., “Adaptive optics development at the German solartelescopes,”Appl. Optics49, G155 (2010).
[27] Hofmann, A., Rendtel, J., and Arlt, K., “Toward polarimetry with GREGOR – testing the GREGOR PolarimetricUnit,” Centr. Eur. Astrophys. Bull.33, 317–325 (2009).
[28] Collados, M., Calcines, A., Dıaz, J. J., Hernandez, E., Lopez, R., and Paez, E., “A high-resolution spectrograph forthe solar telescope GREGOR,” in [Ground-based and airborne instrumentation for astronomy II ], McLean, I. S. andCasali, M. M., eds.,Proc. SPIE7014, 70145Z (2008).
[29] Beck, C., Mikurda, K., Bellot Rubio, L. R., Kentischer,T., and Collados, M., “Multi-wavelength observations at theGerman VTT on Tenerife,” in [Modern solar facilities - advanced solar science], Kneer, F., Puschmann, K. G., andWittmann, A. D., eds., 55 (2007).
[30] Beck, C., Rezaei, R., and Fabbian, D., “Stray-light contamination and spatial deconvolution of slit-spectrographobservations,”A&A 535, A129 (2011).
[31] Lofdahl, M. G. and Scharmer, G. B., “Sources of straylight in the post-focus imaging instrumentation of the Swedish1-m Solar Telescope,”A&A 537, A80 (2012).
[32] Allende Prieto, C., Asplund, M., and Fabiani Bendicho,P., “Center-to-limb variation of solar line profiles as a test ofNLTE line formation calculations,”A&A 423, 1109–1117 (2004).
[33] Cabrera Solana, D., Bellot Rubio, L. R., Beck, C., and Del Toro Iniesta, J. C., “Temporal evolution of the Evershedflow in sunspots. I. Observational characterization of Evershed clouds,”A&A 475, 1067–1079 (2007).
[34] Denker, C. and Tritschler, A., “Measuring and Maintaining the Plate Parallelism of Fabry-Perot Etalons,”PASP117,1435–1444 (2005).
[35] Lites, B. W., Rutten, R. J., and Kalkofen, W., “Dynamicsof the solar chromosphere. I - Long-period network oscil-lations,”ApJ414, 345–356 (1993).
[36] Rouppe van der Voort, L. H. M., “Penumbral structure andkinematics from high-spatial-resolution observations ofCa II K,” A&A 389, 1020–1038 (2002).
[37] Beck, C., Schmidt, W., Kentischer, T., and Elmore, D., “Polarimetric Littrow Spectrograph - instrument calibrationand first measurements,”A&A 437, 1159–1167 (2005).
[38] Beck, C., Schmidt, W., Rezaei, R., and Rammacher, W., “The signature of chromospheric heating in Ca II H spectra,”A&A 479, 213–227 (2008).
[39] Sistla, G. and Harvey, J. W., “Fraunhofer lines withoutZeeman splitting,”Sol. Phys.12, 66–68 (1970).[40] Harvey, J. W., “Fraunhofer lines with large Zeeman splitting,” Sol. Phys.28, 9–13 (1973).[41] de Boer, C. R.,Speckle-Interferometrie und ihre Anwendungen auf die Sonnenbeobachtung, PhD thesis, Universitats-
Sternwarte Gottingen, Germany (1993).[42] Puschmann, K. G. and Sailer, M., “Speckle reconstruction of photometric data observed with adaptive optics,”
A&A 454, 1011–1019 (2006).[43] Denker, C., Deng, N., Rimmele, T. R., Tritschler, A., and Verdoni, A., “Field-dependent adaptive optics correction
derived with the spectral ratio technique,”Sol. Phys.241, 411–426 (2007).[44] Lofdahl, M. G., “Multi-frame blind deconvolution with linear equality constraints,” in [Image reconstruction from
incomplete data II], Bones, P. J., Fiddy, M. A., and Millane, R. P., eds.,Proc. SPIE4792, 146–155 (2002).[45] van Noort, M., Rouppe van der Voort, L., and Lofdahl, M.G., “Solar image restoration by use of multi-frame blind
de-convolution with multiple objects and phase diversity,” Sol. Phys.228, 191–215 (2005).[46] Woger, F., von der Luhe, O., and Reardon, K., “Speckleinterferometry with adaptive optics corrected solar data,”
A&A 488, 375–381 (2008).[47] Woger, F. and von der Luhe, O., “KISIP: a software package for speckle interferometry of adaptive optics corrected
solar data,” in [Advanced software and control for astronomy II], Bridger, A. and Radziwill, N. M., eds.,Proc. SPIE7019, 70191E (2008).
[48] Excelis Visual Information Solutions, “One proposal for an IDL coding standard,” Article ID 4120 (2006).[49] Hanisch, R. J., Farris, A., Greisen, E. W., Pence, W. D.,Schlesinger, B. M., Teuben, P. J., et al., “Definition of the
Flexible Image Transport System (FITS),”A&A 376, 359–380 (2001).[50] Ponz, J. D., Thompson, R. W., and Munoz, J. R., “The FITS image extension,”A&ASS105, 53–55 (1994).
